Echoes from the past: New insights into the early hominin cochlea from a phylo-morphometric approach

(1)

Open Archive TOULOUSE Archive Ouverte (OATAO)

This is an author-deposited version published in :

http://oatao.univ-toulouse.fr/

Eprints ID : 19798

To link to this article : DOI:10.1016/j.crpv.2017.02.003

URL :

http://dx.doi.org/10.1016/j.crpv.2017.02.003

To cite this version : Braga, José and Bouvier, Priscille and Romeyer

Dherbey, Jordan and Balaresque, Patricia and Risser, Laurent and

Loubes, Jean-Michel and Dumoncel, Jean and Duployer, Benjamin

and Tenailleau, Christophe

Echoes from the past: New insights into

the early hominin cochlea from a phylo-morphometric approach.

(2017) Comptes Rendus Palevol, vol. 16 (n° 5-6). pp. 508-520. ISSN

1631-0683

Any correspondence concerning this service should be sent to the repository

administrator:

staff-oatao@listes-diff.inp-toulouse.fr

OATAO is an open access repository that collects the work of Toulouse researchers and

makes it freely available over the web where possible.

(2)

ContentslistsavailableatScienceDirect

Comptes

Rendus

Palevol

ww w . sc i e n c e d i r e c t . c o m

Human

Palaeontology

and

Prehistory

(Palaeoanthropology)

Echoes

from

the

past:

New

insights

into

the

early

hominin

cochlea

from

a

phylo-morphometric

approach

Échos

du

passé

:

éclairages

nouveaux

sur

la

cochlée

des

premiers

homininés

par

une

approche

phylo-morphométrique

José

Braga

a,b,∗

_,

_Priscille

_Bouvier

a,c

_,

_Jordan

_Romeyer

_Dherbey

a

_,

Patricia

Balaresque

a

_,

_Laurent

_Risser

d

_,

_Jean-Michel

_Loubes

d

_,

_Jean

_Dumoncel

a

_,

Benjamin

Duployer

e

_,

_Christophe

_Tenailleau

e

a_{Computer-assisted}_{Palaeoanthropology}_Team,_UMR₅₂₈₈_CNRS,_Université_de_Toulouse_{(Paul-Sabatier),}_37,_allées_J-Guesde,

31000Toulouse,France

b_Evolutionary_Studies_Institute,_University_of_{Witwatersrand,}_1,_Jan_Smuts_Avenue,_Braamfontein_2000,_{Johannesburg,}_South_Africa c_Institut_national_des_sciences_appliquées,_135,_Avenue_de_Rangueil,₃₁₀₇₇_Toulouse_cedex_4,_France

d_Statistics_and_Probability_Team,_Institute_of_Mathematics_of_Toulouse,_UMR₅₂₁₉_{CNRS–Université}_de_Toulouse_{(Paul-Sabatier),}

118,routedeNarbonne,31062Toulouse,France

e_CIRIMAT,_UMR₅₀₈₅_{CNRS–Université}_de_Toulouse_{(Paul-Sabatier),}_118,_route_de_Narbonne,₃₁₀₆₂_Toulouse,_France

a

r

t

i

c

l

e

i

n

f

o

Articlehistory:

Received10October2016

Acceptedafterrevision27February2017 Availableonline18May2017

HandledbyRobertoMacchiarelli andClémentZanolli

Keywords: Cochlea Hominins Phylogram Brownianevolution Australopithecusafricanus Paranthropusrobustus Homo

a

b

s

t

r

a

c

t

Weinvestigatecochlearvariation,anindirectevidenceofauditorycapacitiesamongearly homininsandextantcatarrhinespecies,inordertoassess(i)thephylogeneticsignalof relativeexternalcochlearlength(RECL)andovalwindowarea(OWA),(ii)the evolution-arymodelwiththehighestprobabilityofexplainingourobserveddata,(iii)somehominin ancestralnodesforRECLandOWA.RECLhasahighphylogeneticsignalundera Brown-ianmotionmodel,andiscloselycorrelatedwithbodymass.Ourmodel-basedmethod hastheadvantageoverparsimony-basedmethodsofincorporatingbranchlengthsina phylo-morphospace,andthisshowsRECLshiftedtowardssigniﬁcantlyhighervaluesat theHomoerectus-Homosapiensnode.WealsoobservethattheStW53andKB6067fossil specimensfromSterkfonteinandKromdraailikelyrepresentoneortwodistinct, smaller-bodiedandlessderivedhomininform(s)comparedtoParanthropusspecimensrepresented atSwartkrans.

©2017Acad ´emiedessciences.PublishedbyElsevierMassonSAS.Thisisanopenaccess articleundertheCCBY-NC-NDlicense(http://creativecommons.org/licenses/by-nc-nd/4. 0/). Motsclés: Cochlée Homininés Phylogramme Évolutionbrownienne

r

é

s

u

m

é

Nousexaminonslavariationcochléaire,témoinindirectdescapacitésauditivesdes pre-miershomininésainsiqued’espècesactuellesdecatarrhiniens,aﬁnd’évaluer(i)lesignal phylogénétiquedelalongueurexternerelativedelacochlée(RECL)etdelasurfacedela fenêtreovale(OWA),(ii)lemodèleévolutifmontrantlaplusforteprobabilitéd’expliquer

∗ Correspondingauthorat:Computer-assistedPalaeoanthropologyTeam,UMR5288CNRS,universitédeToulouse(Paul-Sabatier),Toulouse,France. E-mailaddress:jose.braga@univ-tlse3.fr(J.Braga).

http://dx.doi.org/10.1016/j.crpv.2017.02.003

(3)

Australopithecusafricanus Paranthropusrobustus Homo

nosobservations,(iii)certainesvaleursancestralesd’homininéspourRECLetOWA.RECL montreunsignalphylogénétiqueélevésousunmodèlebrownien,maisaussiune corréla-tionétroiteaveclamassecorporelle.Notreméthode,baséesurunmodèleévolutif,présente l’avantagedetenircomptedeslongueursdebranches(contrairementauxméthodes cladis-tiques,baséessurlaparcimonie)dansunespacephylo-morphologiquesoulignantune évolutiondeRECLversdesvaleurssigniﬁcativementplusélevéesuniquementaunœud (Homoerectus,Homosapiens).NousobservonségalementquelesfossilesStW53etKB6067 (provenantrespectivementdeSterkfonteinetdeKromdraai)représententprobablement uneoudeuxforme(s)d’homininé(s)depluspetitetaillecorporelleetmoinsdérivée(s),par rapportàParanthropusreprésentéàSwartkrans.

1. Introduction

Amongsttheﬁvemainsensorysystems,thecochleais theonlyorganthatimprintsdetailsofitsoverallstructure withinbone;this takestheformofaspiral-shaped cav-ityhousedbythepetrosalpartofthetemporal.Atleast twofeaturesofthiscavitycanbedeterminedandenable goodestimatesofthehearingcapacitiesinfossilhominins: therelativeexternalcochlearlength(RECL)andtheoval windowarea(OWA)(seeBragaetal.,2013,2015formore details).Forinstance,cochlearlengthistakenasaproxy measureof ashorter basilarmembrane length(withits sensorstunedtohighfrequenciesatitsbaseand lower frequenciesprogressivelytowardstheapex),andwas sug-gestedtoprovideagoodestimateoflow-frequencyhearing innon-humanprimates(ColemanandColbert,2010).One canthereforequestionhow interactionsof ourhominin ancestorswithenvironmentalsignals(i.e.habitatacoustics andvocalizations)mayhaveplayedaroleintheevolution oftheuniquelow-frequencysensitivitydisplayedonlyby modernhumans(i.e.Homosapienssapiens)among catar-rhineprimates(Coleman,2009).Thisquestionisbeyond thescopeofthepresentstudy.

However,sincecochlearfeaturesareusefulin recons-tructing the evolutionary history of auditory capacities among primates (Coleman and Colbert,2010; Coleman et al., 2010), and show an association with phylogeny (Bragaetal.,2015),weaddresstwoquestionsinthispaper. First, can statistical procedures improve the detection of taxa that deviate signiﬁcantly from general allomet-ricequations(i.e.havelargerorsmallercochlearfeatures giventheirbodysize)?Second,cangrossgeometrical fea-turesofthecochleainanunknownancestralspeciesbe accuratelypredictedfromknowledgeofitsphylogenetic nodal position? In order to address these issues, it is desirable to determine whetherthe cochlear geometri-cal variation observedamong earlyhomininsand other catarrhinespeciesisphylogeneticallymeaningful (similar-itiesindicatingsharedrecentcommonancestry)inpossible relationtobodysize,andwhetherthiscanbetested accord-ingtodifferentexplicitevolutionarymodels(e.g.,Brownian motionversusmodelswithvariableratesofevolution).

In a recent paper, Braga et al. (2015)used microfo-cus X-raycomputed tomography (micro-ct)to measure thestrengthofRECLandOWAphylogeneticsignals,and

to determine whether some hominin species showed cochlearshiftsfortheirbodymassaftercorrectingfor gene-based phylogeny.It was concludedthat RECL evolution in apes occurredmainlythrough body-mass-dependent andnon-homoplasicchanges.Moreover,bothpremodern andmodern humans(Homo erectusand H.sapiens sapi-ens, respectively) showed RECL and OWA values larger than expected for theirbody mass (using phylogeneti-callycontrolledlinearregressions),aconditionnotfound in theirnon-humanhomininpredecessors (Braga et al., 2015).However,inthatstudy,allthephylogeneticanalyses assumedthat Brownianmotionwasthebest evolution-arymodeltoexplainthecochleardataobservedamong catarrhines.In aBrownianmotionmodeloftrait evolu-tion,theexpectedphenotypicdifferencebetweensister speciesgrowsproportionaltothetimesincetheyshared a commonancestor(i.e. thesumof thebranchlengths between the two taxa) (Nunn, 2011). Given the avail-ableevidencethatfunctionalsystemsoftendonotevolve atconstantratesbutinsteadshowstrongpositive selec-tions(withacceleratedevolutionarychanges)(e.g.,Clark etal.,2003),itisnecessarytotestwhetherRECLandOWA mayhaveevolvedfollowinganon-Brownianmodelbefore attemptingtoreconstructancestralvalues.

BothRECLand/orOWAvalueshavebeeninvestigated in Australopithecus (Sts 5, StW 329, StW 98, StW 255), Paranthropus(KB 6067,TM1517,SK879, SKW18)and H.erectus sensu lato(or H.ergaster)(SK 847) (formore details,seeBragaetal.,2013:Table3;Bragaetal.,2015: S1 table). However,since this study focuses mainly on phylogeneticissues,theKB6067specimenistreated sep-aratelyfromtheSwartkransParanthropussamplebecause ithasbeenpreliminarilyinterpretedto“representamore primitiveconditionfortheP.robustuslineage,withmore similarity to some Sterkfontein Member 4 specimens” (Bragaetal.,2013:455).Moreover,inthepresentstudy, weuseonlyfossilspecimens withbothRECL andOWA values,hence allowingcomparisonsofthephylogenetic resultsforthesetwoparametersbyusingthesame sam-ples.Therefore,Sts5andTM1517areexcludedfromour sample.

In addition to published data, the ﬁrst aim of this studyistoprovidefurtherRECLandOWAmicro-ct mea-surements for three early homininspecimens from the Sterkfonteinsite (SouthAfrica).The ﬁrst,StW 498ehas

(4)

beenattributedtothe“Paranthropus-like” Australopithe-cus prometheus, considered as a second species from SterkfonteinMember4byClarke(2008)(formoredetails, seetherecentreviewbyGrine,2013).StW151hasbeen regardedasrepresenting“ahominidmorederivedtowards anearlyHomoconditionthantherestoftheA.africanus samplefromMember4”(Moggi-Cecchietal.,1998;p.462) andStW53 hasbeenattributedtoearlyHomo(Hughes andTobias,1977),toA.africanus(Clarke,2008;Kumanand Clarke,2000)ortoaformthatismorecloselyafﬁliatedto A.africanusthantoearlyHomo(Braga,1998).

Thepresentstudyhasthreeothermainaimsemploying specimensrepresenting9hominoidand13cercopithecoid extantspeciesandtheirassociatedgene-basedconsensus phylogram (i.e. a phylogenetic tree with “molecular-calibrated” branch lengths) obtained using 10kTrees for Primates, V2 (http://www.10ktrees.fas.harvard.edu/) (Arnoldetal.,2010).First,weexplorefurther(ascompared toBragaetal.,2015)thephylogeneticsignalofRECLand OWAinextantspeciesbeforeandaftercontrollingforbody size.Second,weinvestigatewhethertheBrownian evolu-tionarymodel,theOrnstein–Uhlenbeck(OU)process,or theacceleratingversusdeceleratingratesofcharacter evo-lution(ACDC)bestexplainstheobservedRECLandOWA datainthesespecies.Third,weusetheevolutionarymodel thatismostappropriatetoourdata(e.g.,Brownianmotion orOrnstein–UhlenbeckorACDC)toestimatesomehominin ancestralconditionsofRECLandOWAfromaphylogram

thatcombinesdataforextantandfossilspecies(Fig.1;see detailsbelow).

2. Materialsandmethods 2.1. Micro-ctdata

Table1presentsthedetailsofRECL andOWAvalues (given tothenearest one-tenthmillimeter) used inthe presentstudyforeachfossilhomininspecimen,aswellas themeanvaluesforfossilandextantspecies.

Mostofthemicro-ctmeasurementsusedinthisstudy wereemployedpreviouslybyBragaetal.(2015)(seeTable S1formoredetails)withisometricvoxeldimensions ran-gingfrom7.0to41microns(␮m).Therefore,wedonot repeatherethedetailsaboutthesamplesand thebody massdataalreadygiveninBragaetal.(2015).Additional measurements of RECL and OWA were taken from the micro-ctsof7Gorillagorillaskulls(fromspecimenshoused atthe“Muséeroyaldel’Afriquecentrale”,Tervuren, Bel-gium; with only one of them of known sex) obtained using the Nanotom (GE Sensing) at the “Fédération de recherche”FERMAT(CIRIMAT,Toulouse)(allwith isomet-ricvoxeldimensionsof32.6␮m).Themicro-ctdataforStW 498e,StW151andStW53wereobtainedinSouthAfrica usingtheXTH225/320LCdualsourcesystem(Nikon)atthe PalaeosciencesCentre, Universityof theWitwatersrand,

Fig.1. Phylogramfor22extantcatarrhinespeciesandthreeextincthomininspeciesmadefromgraftingthetreefortheextincthomininsusingmorphology ontotheextantmoleculartree.

Fig.1. Phylogrammepour22espècesactuellesdecatarrhiniensettroisespècesd’homininésfossiles,construitengreffantl’arbredesfossilesétabliparla morphologiesurceluiobtenuparlesdonnéesmoléculaires.

(5)

Table1

Micro-ctrelativeexternalcochlearlength(RECL)andovalwindowarea (OWA)meanvaluesforextantandfossilspecies,andmeasurementsfor fossilhomininspecimens.Voxelsizesareindicatedinmicrons.

Tableau1

ValeursmoyennesdeRECLandOWApourlesespècesactuellesetfossiles, etmesuresindividuellespourleshomininésfossiles,obtenuespar micro-tomographies.Lesdimensionsdesvoxelsizessontindiquéesenmicrons.

RECL OWA Voxelsize

Homoerectus SK847(R) 14.2 3.3 21.7 Paranthropus(n=3) 14.3 4.1 SK879(L) 14.8 4.3 9.2 SKW18(R) 13.8 3.9 11.1 KB6067 11.8 2.8 7.4 Australopithecus(n=6) 12.5 2.6 StW329 11.8 2.0 33.1 StW98 12.7 3.1 33.1 StW255 12.8 2.8 33.1 StW498e 13.3 2.0 28.1 StW53g 12.3 2.5 30.5 StW151c 12.2 3.4 28.3 Homosapiens(n=22) 14.6 3.7 41.0 Pantroglodytes(n=9) 13.3 3.1 41.0 Panpaniscus(n=7) 12.5 2.5 8.0–41.0 Gorillagorilla(n=14) 14.6 3.9 32.6–41.0 Pongopygmaeus(n=8) 13.6 4.0 12.0–41.0 Nomascusconcolor(n=1) 9.7 1.6 41.0 Hylobatesmoloch(n=1) 10.2 1.8 41.0 Hylobateslar(n=1) 10.9 1.5 9.0 Hylobatesagilis(n=2) 10.4 1.6 7.3–8.3 Papiohamadryas(n=2) 10.5 1.6 41.0 Papiocynocephalus(n=5) 10.4 1.4 7.8-41.0 Papioursinus(n=1) 11.0 1.6 41.0 Papioanubis.(n=2) 11.8 1.5 8.1-8.6 Mandrillussphinx(n=1) 11.2 1.6 41.0 Macacaradiata(n=1) 9.6 0.9 41.0 Macacasylvanus(n=2) 9.5 1.3 41.0 Cercopithecusmona(n=1) 9.8 1.2 41.0 Cercopithecushamlyni(n=1) 9.1 1.1 41.0 Cercocebustorquatus(n=1) 10.3 1.5 41.0 Colobusangolensis(n=3) 10.2 1.4 7.0–41.0 Colobusguereza(n=4) 10.0 1.3 41.0 Piliocolobusbadius(n=4) 9.8 1.0 41.0

Johannesburg, with isometricvoxel dimensions ranging

from28.1to30.5␮m.

2.2. MeasurementprotocolforOWA

Themicro-ctmeasurementmethodologyofRECLand

OWAisdetailedinBragaetal.(2015).Moreover,we

illus-tratethemicro-ctmeasurementprotocolusedinthisstudy forOWAwithanexampletakenfortheP.robustusSK879 specimen(Fig.2).Thisexample ismotivatedin partby thedifﬁcultiesexpressedintheliteratureaboutaccurate measuresoftheovalwindowinthenarrowovalwindow niche(fenestravertibuli),avariablydeepdepressionlocated partlybehindtheoverhangingandprominentpromontory of thetympanic cavity (promontoriumtympani)(Fig.2). In a detailed analysisof thestapes footplate,Simet al. (2013)stated thatmeasurementstaken onphotographs “might not be the most precise way of measuring the dimensionsoftheossicles”(op.cit.,p.160).Noussiosetal. (2016)alsoconsideredthatphotographicmeasurementsof thestapesfootplateintheovalwindownichemayentail

errorsregardingtheprojectionofsuchasmallsurfaceinto ascreen(parallaxerror).Sinceseveralstudiesnow con-siderinsitumicro-ctmeasurementsofthemiddleearas moreaccurateandefﬁcientthanmeasurementstakenon photographs(thispointwillbediscussedinmoredetails below),weprioritizemicro-ctstomeasureOWAinfossil homininsandextantcatarrhines(Bragaetal.,2013,2015). We ﬁrst visualized (by using the Avizo solftware.

https://www.fei.com/software/amira-avizo/)theoval win-downichein3D afterextracting anisosurfacefromthe micro-ctdataset,asshowninFig.2awiththecaseofSK 879.Wethendefinedanobliqueslicethatbest-fittedthe completeoutlineoftheovalwindowatthelocationofthe annularstapedialligament,aringoffibroustissuethat con-nectstheovalwindowtothestapesfootplate.TheOWA wasthenmeasuredfromitssegmentationonthisoblique slice.Sinceweusedhighresolutionmicro-ctdata(Table1), theoutlineoftheinsertionareaoftheannularstapedial ligamentcouldbevisualizedaccurately.Thisallowedusto reducetoaminimum(lessthan5%)theinterobserver vari-abilitybetweentwooperators(J.B.andJ.R.D.)insettingan obliquesliceintheplaneoftheOWA.

2.3. Themaximumlikelihoodframework

Allthephylogeneticmethodsandtestspresentedbelow havebeenimplementedbyusingthestatisticalmaximum likelihood(ML)framework thatwasﬁrstintroducedfor phylogeneticstudiesbyEdwardsandCavalli-Sforza(1964)

forgenefrequencydata.TheapplicationofMLto phyloge-niesinvolvessearchingforthesingletree(i.e.atopology with branch lengths) that, under a given evolutionary model,bestexplainstheobserveddata(i.e.thetraitsfor eachspecies,formoredetails,seeBaumandSmith,2013). Indoingso,itisconventionaltorecordthelogarithmof thelikelihood,thelog-likelihood(toavoidproblems asso-ciatedwithhandlingverysmallnumbers)ofthattreeas themaincriterion.

Animportantpointtostresshereisthatanalysesfor morphologicaldataarestilloftenbasedonparsimony,a methodthatislesscomputationallydemandingand “con-sistentif therateof evolutionarychange perbranchof thetreeissufﬁcientlysmall”(Felsenstein,2004:122). Intu-itivelyspeaking,thestatisticalpropertyofparsimony(for moredetails,seeFelsenstein,2004)meansthatit mini-mizesthenumberofchangesthathaveoccurredineach branchandthereforeassumesalow rateofchange(i.e. fewerevolutionaryevents,sothephylogramwillshow rel-ativelyshorterbranchlengthsinallcharacters).Moreover, the advantage of ML methods over parsimony analysis is thatthe former useinformation onbrancheslengths availableinphylograms.Giventhefactthatdifferent char-actersareoftenexpectedtoevolveatdifferentratesand thatlongerbranchesoffermoreopportunitiesforchanges to occur than shorter branches, ML methods are par-ticularlyuseful in this context. We use freely available packages (listed below) in the R Statistical Computing environment http://www.r-project.org/. These packages aresummarizedintheCRANproject: http://www.cran.r-project.org/web/views/Phylogenetics.html.

(6)

Fig.2. 3Dreconstructionfrommicro-ctdataofthemiddleearcavityofSK879(A);obliquesliceusedtomeasureovalwindowarea(OWA)andsuperimposed onagridof0.5×0.5mmuniformlydistributedsquares(B);and3Dviewoftheovalwindowniche(C)usingexactlythesameviewpoint,magniﬁcation, gridandorientationasinB.

Fig.2. Reconstruction3Dàpartirdedonnéesmicro-tomographiques,delacavitédel’oreillemoyennedeSK879(A);coupeobliqueutiliséepourmesurer OWAetsuperposéeàunegrillecomposéedecarrésde0,5mmdecôté(B);etvue3Ddelanichedelafenêtreovale(C)enutilisantexactementlesmêmes pointsdevue,grossissement,grilleetorientationquepourB.

2.4. Thephylogeneticsignal

Wefirstinvestigatewhetherthecochleardataobserved forextantspeciesderivefromanonrandom,gene-based tree-likedescentprocess.Thebasicprincipleisto deter-minewhetheragiventreederivedfromgeneticdistances betterfitsasetofcochleardata(asrepresentedbyspecies mean values) at its tips, when compared with the fit obtainedwhenthecochleardatahavebeenrandomly per-mutedacross the tips of the tree, thus destroying any phylogeneticsignalthatmayhaveexisted.Weinvestigate thephylogenetic signal in this datasetbeforeand after controllingfor bodysize.We usedistinctRpackagesto comparetwoestimatesofthephylogeneticsignalforRECL andOWAconsideredseparately:theofPagel(1999)the KofBlomberg(Blombergetal.,2003).

Pagel’s(1999)isaparametermultipliedtoeachofthe off-diagonalelementsof thevariance-covariancematrix (seeNunn,2011,formoredetails).When=0,thetreehas asinglepolytomyatthebasalnodeforallspecies,whereas when=1theoriginalcandidatetreeisrecovered. Statis-ticaltestsforphylogeneticsignalareperformedunderthe nullhypothesisthat=0.Testsforlesssignalthanthe can-didatetreeareperformedunderthenullhypothesisthat =1.WecomputePagel’swiththeRpackage“Geiger”

(itsfunction“ﬁtContinuous”)undertheassumptionof a Brownianmodelofevolution(seebelow).Wealsocompute Pagel’swiththeRpackage“Caper”(itsfunction“pgls”) withnopriorassumptionaboutanygivenmodelof evo-lution.Inallinstances,wetestournullhypotheseswith thelikelihoodratio(LR)test(twicethelog-likelihood dif-ferencebetweenthetwomodelsisexpectedtoﬁtaChi2

distributionwithPdegreesoffreedom)(BaumandSmith, 2013;Nunn,2011).

We compute Blomberg’sK statistic(Blomberg et al., 2003)withtheR“Picante” and“Ape”packagesinorder totestalsowhethertheobserveddistributions forRECL and OWAacrosscatarrhine speciesexhibitmoreor less divergencethanexpectedwhenevolvingunderBrownian motion.ValuesofKrangefrom0toinﬁnity,withK=1 indi-catingBrownianmotionevolution.K>1indicatesthatclose relativesaremoresimilarthanexpected,andK<1 indi-catesmoredivergencebetweentaxathanexpectedunder aBrownianmodel.Inordertodecidewhetherthe calcu-latedKvaluecorrespondstoaBrownianmodel,wetest ifitissigniﬁcantlyhigherthanthemeanKvalueobtained aftersimulating1000datasets(i.e.trees)withrandom per-mutationsofthetips.

We also use phylogenetic generalized least squares (PGLS)(withthe“pgls”functionoftheRpackage“Caper”)to

(7)

assesswhethereitherRECLorOWAconsideredseparately mayhaveundergonecorrelatedevolutionwithbodymass (BM).Therefore,weestimatetheslopesofRECLversusBM andOWAversusBMregressions,andtestwhetherthese slopesaresigniﬁcantlydifferentfrom0.Wesubsequently calculatethephylogeneticsignalofRECLandOWAafter allometriccorrection(i.e.removingtheallometriceffectof bodymasshererepresentedbymeanspeciesvalues). 2.5. Choosinganevolutionarymodel

Inthispartofthestudy,weaimedtocharacterizethe tempo(orrate)andthemode(themechanisms)ofRECL andOWAevolutionarychangesbytestingwhich evolution-arymodelbestexplainstheobserveddata.Weconsiderthe followingthreeevolutionarymodels:BM,OUandACDC.

We ﬁrst consider theBrownian motion(BM) model assumingthatrates ofevolutionare constantover time (i.e.largerchangesmorelikelyoccuronlongerbranches). Therefore,this modelwasalsocalled“constantvariance process”(Freckletonetal.,2002;Pagel,2002).Inthismodel, theinstantaneousphenotypicvariance2_per_unit_time_t

(i.e.thechangealongabranchoflengtht,orrateofchange) isdrawnfromanormaldistributionwithmean0and vari-ance2_t._Therefore,_the_trait_variance_is_proportional_to

time(orbranchlength)andthedisplacementsin differ-entbranchesofatreeareindependent.Itisimportantto addherethattheBMmodeldoesnotnecessarilyindicate neutralorrandomevolution.Itcaninsteadreﬂectadaptive evolutionarychanges(Nunn,2011).

IncontrasttoBM,theOrnstein–Uhlenbeck(OU)model can beviewedasacting tolimitevolutionary variation. Indeed, it isbiologicallyunrealistic toconsider thatthe rangeofatraitvaluecouldvaryinﬁnitely.TheOUmodelcan thereforebeapplicablewhentraitslikelyevolvedwith con-straintsontheirmaximumorminimumvalues,depending onthestrengthofstabilizingselection(Felsenstein,1988; Garlandetal.,1993).ItisaBMmodelpulledtoincludeone ormoreselectiveoptimathatassertanattractiveforceon Browniantraitevolution.Whenthestrengthofthis attrac-tioniszero,theOUmodelisidenticaltoaBMmodel.Very strongstabilizingselectioncanobliteratetheeffectsof his-torysuchthatphylogeneticsignaldisappears.Inthisevent, trait values willbe more similarin closely ordistantly relatedspeciesexperiencingthesameselectivepressures. Theaccelerationor deceleration(ACDC) ofBrownian motion evolution from the root to the tips of the tree is thethird model usedin this study. TheACDC model (alsocalled“earlyburst”)describesevolutionthateither increases(accelerates,AC)ordecreases(decelerates,DC) inrateovertime.

Wechoosetheevolutionarymodelthatbestdescribes thespeciesdatasampledinourstudywiththeuseofLR testsandtheAkaikeinformationcriterion(AIC)(Akaike, 1974).ThemodelwiththesmallestAICisthepreferredone. WealsousethecorrectedAIC(AICc)forsmallsamples. 2.6. Estimatesofancestralvalues

We estimate RECL and OWA ancestral values (using the preferred evolutionary model chosen in the

previousstep) at internal nodes of a phylogeny recon-structed with morphological and molecular data by grafting a tree for three extinct hominin species of known cochlear values (A. africanus, P. robustus and H.erectus)onto thegene-basedphylogram representing contemporaneouscatarrhinespecies(Fig.1).Theextinct homininphylogram(i.e.branchlengthsproportionaltothe amountofevolutionarychange)isobtainedfromOrgan etal. (2011)and wasinferred using109 morphological characters from Strait and Grine (2004) in a Bayesian framework.

WeusetheRpackage“Phytools”(andits“fastAnc” func-tion)(Revell,2012)tocomputeancestralvaluesand95%CIs (toreflectuncertainty)usingML(i.e.findingvaluesthat maximizetheprobabilityofourdata).We considerthat ourapproachisunbiasedbecauseweusethe evolution-arymodelthatbestfitsourdata.Theabilityofdifferent methods (e.g.,ML under different modelsof evolution) toaccuratelyestimateancestralvaluesand their associ-atederrorshasbeeninvestigatedbyMartins(1999),who showedthataBMmodelperformedreasonablywell, espe-cially for more recent ancestors. Therefore, in order to reduceuncertainty, we limit ourestimates for hominin ancestralnodesthatareclosetothetips(asopposedto thosethataredeeperinthetree).Weconsiderthe follow-ingfournodes:(Pan,Hominins),(A.africanus,P.robustus, Homo), (P. robustus, Homo) and (H. erectus, H. sapiens) (Fig.1).

3. Results

3.1. VariationofRECLandOWAamongearlyhominins Weﬁrstrepresentinasimple2D-pointgraph,the varia-tionofRECLversusOWAamongearlyhominins,H.sapiens sapiens,GorillagorillaandPantroglodytes/paniscus(Fig.3). Atﬁrstglance,weobserveaclearseparation betweena groupcomprisingthefourAustralopithecusspecimensfrom Sterkfontein(StW498e,StW 255/259, StW98 andStW 329),StW53,StW151andKB6067,andasecondgroup madeupofP.robustusfromSwartkrans(SK879andSkW 18)andH.erectus(SK847)withnoticeablyhighervaluesfor bothRECLandOWA(Fig.3).Inthisregard,itisimportantto notethatRECLisabetterdiscriminatorthanOWA,theSK 847OWAvaluefallingwithinthevariabilitysampledfor specimensderivedfromSterkfonteinMember4.Wealso observethatH.sapiensandGorillagorillaspecimensoverlap oneanotherandthatmostofthemhavehighervaluesof RECLthanthatforPantroglodytes/paniscus.Whilethethree specimensofP.robustusfromSwartkransandH.erectus groupfallwithinH.sapiensandGorillagorillaranges,allthe otherfossilhomininssampledherefromKromdraaiand SterkfonteingroupwithvaluesforPantroglodytes/paniscus. Moreover,theStW329specimenshowsremarkablylow RECLandOWAvalues.

3.2. PhylogeneticsignalofRECLandOWA

ForbothRECLandOWA,weobtainahighPagel’sthat issigniﬁcantly differentfrom0 butnotdifferentfrom1,

(8)

Fig.3.2D-pointgraphrepresentingrelativeexternalcochlearlength(RECL)andovalwindowarea(OWA)valuesinfossilhominins,Homosapiens,Panand Gorilla.

Fig.3. Graphique2DdenuagedepointsreprésentantlesvaleursdeRECLetOWAchezleshomininésfossiles,Homosapiens,PanetGorilla.

Table2

PhylogeneticsignalsasrepresentedbytheofPagelandtheKofBlombergforrelativeexternalcochlearlength(RECL)andovalwindowarea(OWA), beforeandaftercorrectionforbodymass(BM),underastandardBrownianmotionmodelofevolution.Theresultsofthelog-likelihoodratiotests,the permutationtests(Kmean)andtheAkaikeinformationcriterion(AIC)(withthecorrectedAIC–AICcforsmallsamples)arealsodetailed.MLformaximum likelihood.

Tableau2

SignauxphylogénétiquesreprésentésparlesdePageletKdeBlombergpourRECLetOWA,avantetaprèscorrectionallométriqueparlamassecorporelle (BM),sousunmodèled’évolutionbrownienstandard.Lesrésultatsdestestsderatiodelog-vraisemblances,destestsdepermutation(Kmean)etcritères d’Akaike(AIC)(valeurAIC–AICccorrigéepourlespetitséchantillons)sontégalementdétaillés.MLpourmaximumdevraisemblance.

RECL OWA BM

ofPagel(ML,Brownian,Geiger) 0.926133 1.000000

ofPagel(ML,Caper) 0.926000 1.000000 1.0000 Log-likelihoodratio 40.089887 – 18.966384 – – AIC –74.179773 – –31.932767 – – AICc –72.846440 – –30.599434 – –

P-value5%(differentfrom0)(Geiger) 2.060154e-05 4.523477e-07

P-value5%(differentfrom0)(Caper) 2.0602e-05 4.5236e-07 4.8937e-06

P-value5%(differentfrom1)(Geiger) 0.1446526 1

P-value5%(differentfrom1)(Caper) 0.14465 1 1

KBlomberg 0.9848426 1.522079 1.229131

Kmean(1000) 0.0001673756 0.001036746

P-value 0.000999001 0.000999001

Pagel:BM(Caper) 0.320 0.735

P-value5%(differentfrom0) 0.10004 0.0003707

(9)

eitherundertheassumptionofaBrownianmodelof

evo-lution,orwithnopriorassumptionaboutanygivenmodel

ofevolution(Table2).Thesametrendisobservedwhen

we computetheBlomberg’s Kstatistic.In all instances, theKvaluesaresigniﬁcantlyhigherthanthoseobtained randomly and indicate either Brownian motion evolu-tion(RECL;K=0.98)orcloserrelativesmoresimilarthan expected(OWA;K=1.52)(Table2).

Since BM shows a highphylogenetic signal for both Pagel’s(signiﬁcantlydifferentfrom0,butnotfrom1)and Blomberg’sK(1.22)(Table2),itisimportanttoinvestigate whetherRECLandOWAstillshowahighphylogenetic sig-nalaftercontrollingforBM.WeobtainalowPagel’s(0.32; notsigniﬁcantlydifferentfrom0)forRECLafter control-lingforBM(r2₌_0.78,_P_<_0.005)._For_OWA,_after_controlling

forBM(r2₌_0.68;_P_<_0.005),_the_Pagel’s_is_higher_than_for

RECL,withavalue(0.735)signiﬁcantlydifferentfrom0,but notdifferentfrom1at0.03%(Table2).

3.3. ChoiceofanevolutionarymodelforRECLandOWA We ﬁrst compute the log-likelihood for a Brownian modelof evolutionforRECL andOWAconsidered sepa-rately(Table3).Wethencomputethelog-likelihoodsfor OUand ACDCmodelsofevolutionof RECLand OWAin ordertodeterminewhetheroneofthembetterexplains ourobserveddatathanBrownianmotion.BothLRtestsand AICcriteria(eitherAICorAICc)showthattheevolutionary modelthatbestexplainsthedata(i.e.RECLandOWAvalues andthegene-basedphylogram)observedinextantspecies isBrownianmotion(Table3).Thismodelisthereforeused forancestralstatesreconstructions.

Table3

Log-likelihoods,LRtestsandAIC(AICc)criteriaforthreedistinctmodels ofevolutionforrelativeexternalcochlearlength(RECL)andovalwindow area(OWA):Brownianmotionmodel,Ornstein–Uhlenbeckmodeland ACDCor“earlyburst”model.

Tableau3

Log-vraisemblances, testsderatio de log-vraisemblances, etcritères d’Akaike(AICetAICc)pourtroismodèlesd’évolutiondistinctsdeRECLet OWA:modèlebrownien,modèled’Ornstein–UhlenbecketmodèleACDC ouearlyburst.

RECL OWA

Brownianmotionmodel

Log-likelihood 39.026010 18.966384 AIC –74.052020 –33.932767 AICc –73.420441 –33.301188 Ornstein–Uhlenbeckmodel Log-likelihood 39.076838 18.966384 AIC –72.153675 –31.932767 AICc –70.820342 –30.599434 P-value 0.7498513 1

ACDCor“earlyburst”model

Log-likelihood 39.026003 19.583036

AIC –72.052007 –33.166072

AICc –70.718674 –31.832739

P-value 1 0.2667654

3.4. Ancestralstatesreconstructions

Since both RECL and OWA show high phylogenetic

signals (when considered separately), we estimate the

ancestralvalues(with95%CIs)ofthesetwoparametersat

thethreehomininnodesinthephylogram(Fig.1),aswell

asatthePan-Homininnode(Figs.4and5).

WithregardtoRECL,thereisacleartrendforanincrease oftheancestralvaluesfromtheoldestnode(Pan,Hominins)

Fig.4.Ancestralvaluesforrelativeexternalcochlearlength(RECL)estimatedatthe(Pan,Hominins)(blue),(A.africanus,P.robustus,Homo)(red),(P.robustus, Homo)(yellow)and(H.erectus,H.sapiens)(violet)nodes.Theconﬁdenceintervalsareillustratedwithonestandarddeviation(withtwostandarddeviations intransparency).

Fig.4. ValeursancestralesdeRECLestiméesauxnœuds(Pan,Hominins)(bleu),(A.africanus,P.robustus,Homo)(rouge),(P.robustus,Homo)(jaune)et (H.erectus,H.sapiens)(violet).Lesintervallesdeconﬁancesontillustrésavecunedéviationstandard(avecdeuxdéviationsstandardentransparence).

(10)

Fig.5.Ancestralvaluesforovalwindowarea(OWA)estimatedatthe(Pan,Hominins)(blue),(A.africanus,P.robustus,Homo)(red),(P.robustus,Homo) (yellow)and(H.erectus,H.sapiens)(violet)nodes.Theconﬁdenceintervalsareillustratedwithonestandarddeviation(withtwostandarddeviationsin transparency).

Fig.5. ValeursancestralesdeOWAestiméesauxnœuds(Pan,Hominins)(bleu),(A.africanus,P.robustus,Homo)(rouge),(P.robustus,Homo)(jaune)et (H.erectus,H.sapiens)(violet).Lesintervallesdeconﬁancesontillustrésavecunedéviationstandard(avecdeuxdéviationsstandardentransparence).

totheyoungest oneand (H. erectus,H.sapiens)(Fig.4). The(A.africanus,P.robustus,Homo)node’sCIat95%ﬁts completelywithinthe(Pan,Hominins)node’s CIat95%. Moreover,the(P.robustus,Homo)and(H.erectus,H. sapi-ens)nodes’RECLestimationsfalloutsidethe(A.africanus, P.robustus, Homo) node’s CI at 63% (i.e. withonly one standard deviation) and 95%, respectively. There is no overlap between the (A. africanus, P. robustus, Homo) and the (H. erectus, H. sapiens) nodes’ CIs at 63% (Fig.4).

ThereisalsoatrendforanincreaseoftheOWA ances-tralvalues fromthe oldestnode(Pan, Hominins)to the youngestone(H.erectus,H.sapiens)(Fig.5).AsforRECL,the (A.africanus,P.robustus,Homo)node’sCIat95%is encom-passedbythe(Pan,Hominins)node’sCIat95%.However, thereis moreoverlapbetweenthethree OWAhominin nodes’CIsat63% (Fig.5)than forRECL (Fig.4).In par-ticular,thereisanalmostcompleteoverlapbetweenthe (A.africanus,P.robustus,Homo)andthe(H.erectus,H. sapi-ens)nodes’CIsat63%(Fig.5).

InordertoevaluatetheimpactonestimatesofRECL andOWAancestralvalueswhengraftingatreeforthree extincthomininspeciesontothegene-basedphylogram representingextantspeciesonly,wecomputedthe(Pan, Homo)nodeCIsbeforetheincorporationoffossildata(asin

Bragaetal.,2015).ForbothRECLandOWA,the(Pan,Homo) nodes’CIsat95%(12.5–14.7mmand2.6–3.9mm2_,

respec-tively)werealmostidenticaltothe(H.erectus,H.sapiens) nodes’CIsobtainedfromthetreecombiningextantand fossilspecies(Figs.4and5).Thisresultdemonstratesthe usefulnessofgraftingfossiltipdataontothegene-based phylogenetic treefor ancestralreconstructions at some nodes.

4. Discussion

In order to better interpret our fossil cochlear parameters,weaimedtouseourdatasetrepresenting con-temporaneouscatarrhinespeciesto(i)explorefurtherthe phylogeneticsignalbeforeandaftercontrollingforbody size,(ii)investigatewhethertheBrownianmodelis appro-priatewhenmodelingcochlearevolutionand(iii)estimate somefossilhomininancestralconditionsofRECLandOWA fromaphylogramcombiningextantandfossildata.Here, wediscussonlythetaxonomicandphylogeneticaspects ofourresultssincedirecttranscriptionsoffossilhominin RECL andOWAdifferences intohearingcapabilities and communicationsignalswouldbeprematureforthe follow-ingreasons.Indeed,previousstudiesinlivinggreatapes havedemonstratedthecomplexityofcallingbehaviorson primatesoundandvocalrepertoiresas,forinstance,within highlyvariablecalltypes(e.g.,longvocalizations) varia-tionsinacousticstructure,callrepertoiresizecorrelated withgroupsizeandsocialbonding,sociallylearnedcalls anddivisionsofcallsaccordingtohomerange(Harcourt and Stewart, 2001; Hardus et al., 2009;Marshall et al., 1999).

4.1. RECLandOWAestimatesinfossilhominins

Apart from Braga et al. (2013, 2015), we could not ﬁndanyotherRECLmeasurementsinearlyhomininsto comparewithours.Moreover,wediscussbelowwhywe considerthatthemicro-ctOWAmeasurementspresented here are not interchangeable with published measure-mentsoftheovalwindoworstapesfootplateareatakenon scaleddigitalimagesinfossilhominins(e.g.,Quametal.,

(11)

2015).Quametal.(2015)assessedphotographicallythe OWAorthestapesfootplateareaandconsideredthat,in allinstances(i.e.forallfossilspecimens),thelatter repre-sented90%oftheformer.Ifweusethis90%proportion tocomparethemicro-ctOWAvalues infossilhominins presentedinthisstudy(Table1)andthephotographic mea-surementspublishedinQuametal.(2015,Table1)(e.g., thestapesfootplateareaofSKW18ortheOWAofSK879), weobservemarkeddiscrepancies(exceptforSTW328),in particularforP.robustus.Herewetakeoneexample pre-sentedinFig.2inordertobrieﬂydiscussthispoint.The mostimportantdiscrepancy(1.6mm2₎_is_observed_for_the

SK879specimenillustratedwithanobliqueslicethat best-fitsthecompleteoutlineofitsOWA(Fig.2B)andwitha3D viewofitsovalwindowniche(Fig.2C).Asshownina com-parisonbetweenFig.2BandC,theconfigurationoftheoval windownicheposestwomaindifficultieswhentryingto measurethesmallOWAwithnoparallaxerror.

Theﬁrstdifﬁcultyisduetothefactthat,evenwhenthe middleearcavityisvisuallyaccessiblewithoutendoscopy (asfor SK879),the promontoryof thetympanic cavity preventsthevisualizationofthecompleteoutlineofthe ovalwindowtoallowforanaccuratephotographic mea-surementof OWAtothenearest one-hundredth square millimeterwithnopotentialparallaxerror.Thispotential parallaxerrorexplainswhyinsitumicro-ctmeasurements ofmiddleearfeatureshavebeenconsideredasmore accu-ratethanphotographicmeasurements(e.g.,Noussiosetal., 2016;Simetal.,2013).

Theseconddifﬁcultyisrelatedtotheneedfora per-fectcalibrationofthephotographtobeusedforaccurate measurementsofOWA.First,thescaleusedforcalibration needstobeplacedatpreciselythesamedistancefromthe cameralensastheovalwindowitself.Moreover,ifa mil-limeterrulerscaleisused,itneedstoﬁtinsidethenarrow ovalwindownichenexttotheOWA(Fig.2C),andalsoto lieexactlyinitsplane(i.e.theOWAandtherulerscale needtobeco-planar).Smallmisalignmentand/or place-menterrorsoftherulerscalewillresultinuncertaintieson thecalibrationofthephotographusedformeasurements. InthecaseoftheSK879micro-ctdata,weobtaineda 4.3-mm2_OWA_value₍_Table₁_)._In_Fig.₂_B,_we_superimpose

agridof0.5×0.5mmuniformlydistributedsquares(with surfaceareasof0.25mm2_each)_over_the_oblique_slice_that

weusedtoobtainthisOWAvalueforSK879.Fig.2Cshows exactlythesameviewpoint,magniﬁcationandorientation ofthisspecimenasinFig.2B.Thesamegridis superim-posed ona 3Dendo-tympanal viewof theSK879oval window niche. Avisual comparison shows a noticeable differencebetweenFig.2BandC in themeasure ofthe OWAwithatleast0.5mm2 ₍₂_squares)._As_shown_here,

suchadifferenceisduetotheoverhangingpromontoryof thetympaniccavitythatcouldbealsoseeninaphotograph orunderamicroscope.Moreover,thegridsuperimposed ontheOWAoftheSK879specimen(Fig.2B)allowsthe readertovisuallyassessthatitsvalueiscloseto4.0mm2_.

A 2.43-mm2 _value _was_given _in _Quam _et_al. ₍₂₀₁₅₎ _for

theSK879estimatedstapesfootplatearea.Ifweassume thatthestapesfootplaterepresented90%oftheOWAin SK879(assuggestedinQuametal.,2015),weobtaina 2.7-mm2 _value._This_measurement _is _incompatible_with

themicro-ctvalue(4.3mm2₎_that_we_repeatedly_obtained

forSK879aftercarefulsegmentations(Fig.2).

Anotherimportant pointtodiscussisthe correspon-dencebetweenthestapesfootplateareaandtheOWA.The differencebetweenthesetwoareasisduetothesizeof theannularligamentofthestapes.Inadetailedmicro-CT measurementstudyoftheannularligamentofthehuman stapes,Mohammadietal.(2016)foundithighlyvariablein itsvolumeandthickness.Becauseofthisvariabilityamong modernhumans,wewouldurgecautionwhenestimating OWAfromthestapesfootplateareainotherextantspecies, aswellasinfossilhomininswithasingleconstant pro-portion.Theproportionof90%usedforfossilhomininsin

Quametal.(2015)likelyrepresentsonlytheupperpart ofthevariationoftherelationshipbetweenOWAandthe stapesfootplatearea.Here,wearguethatmore compara-tivemicro-CTmeasurementsoftheOWAandthestapes footplateareaare needed in both modernhumans and otherextantcatarrhinesforbetterestimationsof measure-mentserrorswhenassessingtheformervariablefromthe latter.Indeed,fromthemicro-ct dataofthemiddleear obtainedbyoneof us(J.B.), we observea76–87% vari-ability(causedby interindividualdifferencesand lateral asymmetry)intheproportionoftheOWArepresentedby stapesfootplateareainonly12measurementsformodern humans.Thisobservationiswellinlinewiththeresults presentedinMohammadietal.(2016).Moreover,when weinvestigatetheproportionoftheOWArepresentedby thestapesfootplateareausingourmicro-ctdatafor11 specimensrepresentingnon-humanprimatespecies,we observea rangebetween79%(Papiohamaryas)and90% (Pongopygmaeus).

4.2. Exploringnewbasicranialfeatures

Thepresentstudyconﬁrms,usingadditionaltests,the veryhighphylogeneticsignalofRECLandOWA parame-tersandtheirclosecorrelationwithbodymass,asalready observed by Braga et al. (2015). Moreover, we newly observethataBrownianmotionmodelisthebest evolu-tionaryframeworktocompareRECLandOWAdifferences between catarrhine species, including fossil hominins. Importantly,whileRECLevolutionisassociatedwith evo-lutionary changes in body mass, OWA likely evolved moreindependently.Therefore,comparisonsbetween fos-sil hominin specimens and species based on RECL and OWAmayleadtoreliableidentiﬁcationsofmonophyletic groupswiththeirassociatedsynapomorphies(asopposed tohomoplasies).

Thefossilcochlearvaluespresentedhereconﬁrmthe RECL distinctiveness of theParanthropus and H.erectus specimensfoundatSwartkranscomparedtothe Australo-pithecusspecimensfromSterkfontein(Bragaetal.,2013, 2015).TheParanthropus and H. erectusRECL values are similartomeasurementsforextanthumansandgorillas, butalsonoticeablyhighertothoseforanyspecimenfrom Sterkfontein,includingStW151orStW53(Fig.3). Interest-ingly,theStW53RECLandOWAvaluesappearveryclose tothoserecentlypublishedforKB6067,ajuvenilefossil homininfromtheSouthAfricansiteofKromdraai(Braga

(12)

etal.,2013,2016)whichhasbeenattributedtoan evolu-tionarylessderivedP.robustusformthanthatrepresented atSwartkrans.Indeed,asnotedbyBragaetal.(2013),StW 53andKB6067alsoappearsimilarinthemorphologyof theirsemi-circularcanals.Interestingly,whileStW53and StW151petrousboneshavebeenregardedasveryclose inmorphologyandmorelikethemodernhumancondition thanexhibitedbyotherSterkfonteinspecimens,theOWA sizeinStW53fallsintherangeofP.paniscuswhereasthe muchlargerOWAofStW151fallsin therange ofboth H.sapiensandG.gorilla.Moreover,theRECLvaluesofboth StW53andStW151fallwellbelowtheP.robustus.H. erec-tusandH.sapiensvalues(Fig.3).

As indicated in Braga et al. (2013), the taxonomic attributionof KB6067fromKromdraaiMember3 isas yetunclear.Aﬁrmertaxonomicattributionoftheoldest Kromdraai paranthropines (including KB 6067) is still pending on the analysis of the larger hominin sample discoveredin theoldestfossiliferous deposits fromthis site (Braga and Thackeray, 2016; Braga et al., 2017). This will allow us to determine whether the southern AfricanParanthropushypodigmrepresentseitherasingle and variable P. robustus species with a time span yet unknown, or distinct taxa (i.e. at most the more ple-siomorphicP.robustusfromKromdraaiandP.crassidens fromSwartkrans;atleastasingleevolutionaryP.robustus species consisting in a lineage of ancestral descendant populations;formoredetails,seeBragaetal.,2016,2017). Furtherdiscoveriescombinedwithprogressinmeasuring thecomplexcochlearshapewillallowustotestwhether thevariableOWAandRECLsignaturesobtainedbetween Swartkrans(e.g.,SKW18andSK879)andKromdraai(e.g., KB6067)paranthropinesareconsistentwith“a plesiomor-phicstatusoftheKBhominins,indicativeoftheirancestral statusfortheP.robustus+boiseiclade,givingrisebefore 2.3MatothesplitofP.boiseiinEastAfricaandP.robustus survivorsinSouthAfrica”(Bragaetal.,2016:65).

Eventhoughourresultsarepreliminary,giventhe phy-logenetictestspresented hereandtheiradvantageover parsimony-based interpretations (see below), we argue that StW 53 and KB 6067 may represent one or two distinctsmaller-bodied,lessderivedhomininform(s),as comparedtoParanthropusandH.erectusspecimens repre-sentedatSwartkrans.Weacknowledgethatlabyrinthine similaritiesarenotsufficienttoestablishfirmtaxonomic affiliationsandthatourconclusionsneedtobetestedby furthercomparativemorphologicalstudies.For instance, althoughthespiraledshapeofthecochleaappears glob-allysimilaramongcatarrhines,thedevelopmentofnew 3Dmorphometricmethodsmayuncovertaxonomic dis-tinctionsbecauseitsspiraledstructuremakesitinefficient tostudywithstandardEuclidianmetrics.

The base of the cranium has often been considered evolutionarilymoreconservativethatthefaceandteeth (Bosma,1976;deBeer,1937;Liebermanetal.,1996;Strait, 1998;Straitetal.,1997).Variationsinthecranialbaseother thanthoseinvestigatedhere(e.g.,theangleofthepetrous partofthetemporalbonetothecoronalplane)havebeen consideredusefulin phylogeneticanalyses(Dean,1986; DeanandWood,1981),buthavenotbeentestedusingthe statisticalmethodsemployedhere.Inlinewithourresults,

thesebasicranialfeatureshavebeenregardedasdistinct betweenthecontemporaneousgreatapesand Australop-ithecus,ontheonehand,andParanthropusandHomoon theother.Furthercomparativedataandinvestigationsat both intra-and inter-speciﬁclevelsin extantand fossil taxawillimproveourphylogeneticinterpretations,andin particular, we need data on the cochlear variability in highlydimorphichominidspeciessuchasG.gorilla.Asyet, weonlyinvestigatedtheRECLvariabilityinaG.gorilla sam-pleofverylimitedsize(n=14)andwithmostspecimens (n=12)ofunknownsex(n=12)andtwomales.

4.3. ParanthropusandHomosharedfeatures

On thebasis ofthehighphylogeneticsignalof RECL and OWA and our evolutionary model-based phyloge-netic statistics, here we interpret the P. robustus and Homosharedand derivedcochlear featuresreportedon Swartkrans specimens (i.e. higher values for both RECL and OWA) as synapomorphies. From our very limited sample of fossilhominins and the resultspresented in

Figs. 4and 5,it is temptingtosuggesta A. africanusto P. robustus ancestral-descendant relationship. However, such a conclusionwould be unwise if it rests only on highRECLandOWAphylogeneticsignals.Wenevertheless arguethattheimprovementofourunderstandingofthe phylogeneticrelationshipsbetweenthesouthernAfrican paranthropines and otherearlyhominintaxa, including earlyHomoandParanthropusfromEastAfrica,dependson furtheranalysesoftheskullbase,includingthe morphol-ogyofthecochlea.

Asyet,thecranialbaseﬂexion hasgarnereda lot of attentionand, tothebestof ourknowledge, itsderived morphologysharedbetweenHomoandParanthropushas not been explored among fossil and extant primates usingphylogeneticallyinformedstatisticalanalyses.Even though we used such analyses in the present cochlear study, we cannot yetdetermine whethertheRECL and OWA derived features shared between P. robustus and HomoatSwartkransarealsosharedwitheasternAfrican Paranthropus and Homo. There is no consensus on the basalspeciesleadingtoParanthropusandHomo, depend-ingonthephylogeneticroleaccordedtoA.africanusand P.aethiopicus (Strait and Grine,2004)and Paranthropus phylogeny(WoodandConstantino,2007,ﬁg.10). There-fore, more comparative studies onAustralopithecus and ParanthropusspecimensfromsouthernandeasternAfrica willbeneededtodecidewhetherA.africanusrepresents themostlikelycommonancestorofP.robustusandHomo, ParanthropusandHomoorifanalternativescenarioismore likely.

In linewithresultsfrom Bakeret al.(2016) andthe positiveselectionassociatedwithshiftsinmolararea rela-tivetobodysizeinParanthropus,ourphylogeneticﬁndings supportasigniﬁcantroleofselectioninshapingthe evo-lutionofP.robustusandHomoauditorycapacities,atleast asrepresentedbyRECL(Fig.4).Thepositiveassociation between BMand RECL observed in this study(see also

Braga et al., 2015)meansthat there is an evolutionary correlationbetweenBMand cochlearsizeamong catar-rhines,includingfossilhominins.SinceRECLdoesnotshow

(13)

a phylogeneticsignalsignificantlydifferentfrom0after controlling for BM, the ancestral values proposed here highlight,atleastpartly,theroleofselectiononalarger body size in shaping cochlear evolution in Homo. Only furthermeasurementsofRECLinfossilHomospecimens willallowustodeterminetheexactshiftsincochlearsize relativetobodysizeduringtheearliest partofthe evo-lution ofourgenus.The useofa phylogram combining extantandfossilspeciestoestimateancestralRECL val-uesatthreehomininnodesinthisstudy(Fig.4),confirmsa previousfindingthatRECLis“hypertrophied”inthegenus Homo, asopposed to theaustralopiths (Australopithecus andParanthropus)(Bragaetal.,2015).Indeed,whenRECLis representedinaphylo-morphospace(Fig.4),the(H. erec-tus,H.sapiens)nodes’value isclearly shiftedtowardsa higher RECL, as compared totheancestral valueat the (P.robustus,Homo)node.Moreover,whenweconsiderthe OWAparameterwithitshighphylogeneticsignalafter cor-rectingforBM(contrarytoRECL),weobservethatOWA valuesareshiftedtowardshighervaluesatthe(P.robustus, Homo)node.Furtherestimatesofancestralnodevalueswill beneededtohelprefiningthepredictionsbyaddingnew fossilhomininspeciesvaluesatthetipsofthetreeusedin thestatisticalanalyses.Tothisend,cochlearinvestigations onA.afarensis,P.aethiopicusandP.boiseiwillbecrucial. 4.4. Cladisticversusphylogeneticmethods

Thepresentstudydemonstratesthatasimple Brown-ianmodelof evolutionbestexplainstheobserved RECL andOWAdataincontemporaneousspecies.Therefore,this modelassumingconstantratesofevolutionovertimeis appropriateforancestralreconstructions.Moreover,this resulthighlightstheimportanceofinvestigatingcochlear changesamongfossilhomininsbyusinganevolutionary frameworkwithstretchedandcompressedbranchlengths thatconformstoourassumedunderlyingBrownianmotion mode.Such anevolutionaryframeworkis morerealistic becauseitassumesthatchangesin charactervalues are morelikelyalonglongerbranches,i.e.overlongerstretches oftime.Whenbranchlengthsareavailable,ourproposed model-based method has the advantage of incorporat-ingthesedata.Bycontrast,theparsimony-basedmethods used for cladisticsreconstructions assume that changes areequallylikely onallbranchesof thetree,regardless of theirlength.Thisassumptionwould besafeonly for veryslowlyevolvingtraits,butnotforcochlearevolution amonghominins.Inthiscontext,methodsusing parame-terswithameasuredphylogeneticsignal(i.e.statistically testedusingaphylogeneticcomparativeapproach)willbe veryusefultoinvestigatefurtherthehypothesized Paran-thropusmonophyly(seereviewsinWoodandBoyle,2016; WoodandConstantino,2007),itsdiversiﬁcationinto east-ernandsouthernAfricanformsfromacommonancestor oftensetduringthe2.7–2.3Maperiod,anditsphylogenetic relationshipswiththeoriginofourowngenusHomo. Acknowledgments

Wetakethisopportunitytoshareourmemoriesof Lau-rent Puymerail who was a magniﬁcent and resourceful

youngcolleague.Laurent’scontributionsandpersonality willremainwithusfora longtime tocome.We thank RobertoMacchiarelliandClémentZanollifortheir invita-tiontocontributetothisthematicissueinhonorofLaurent Puymerail’swork.

ThisworkwassupportedbythetheFrenchMinistry of Foreign Affairs, the “Centre national de la recherche scientiﬁque”andits“Projetexploratoirepluridisciplinaire inter-instituts”PEPII-AUDEVOinFrance,theSouthAfrican NationalResearchFoundation.WethankEmmanuel Gilis-senfromthe“Muséeroyaldel’Afriquecentrale”(Tervuren, Belgium),BernhardZipfelandtheFossilAccessCommittee, fromtheUniversityoftheWitwatersrand(Johannesburg, SouthAfrica)forgivingusaccesstothespecimensnewly investigated inthis study. We arealso gratefulto Kris-tianCarlsonand JakataKudakwashefor theirhelpwith the scanning of three fossil hominin specimens from Sterkfontein (South Africa). These data were produced atthePalaeosciencesCentreMicrofocusX-rayComputed Tomography (CT)Facility at the University of the Wit-watersrand.WeﬁnallydeeplythankFrederickGrine,one anonymousreviewerandRobertoMacchiarelli(associate editor)fortheirveryhelpfulcomments,whichimproved themanuscript.

References

Akaike,H.,1974.Anewlookatthestatisticalmodelidentiﬁcation.IEEE Trans.Autom.Control.19,716–723.

Arnold,C.,Matthews,L.J.,Nunn,C.L.,2010.The10kTreesWebsite:anew onlineresourceforprimatephylogeny.Evol.Anthropol.19,114–118. Baker,J.,Meade,A.,Pagel,M.,Venditti,C.,2016.Positivephenotypic

selec-tioninferredfromphylogenies.Biol.J.Linn.Soc.118,95–115. Baum,D.A.,Smith,S.D.,2013.Treethinking.Anintroductionto

phyloge-neticbiology.RobertsandCo,GreenwoodVillage.

Blomberg,S.,Garland,T.,Ives,A.,2003.Testingforphylogeneticsignal incomparativedata:behavioraltraitsaremorelabile.Evolution57, 717–745.

Bosma,J.F.,1976.DevelopmentoftheBasicranium.NationalInstitutesof Health,Bethesda,Maryland.

Braga,J.,1998.Chimpanzeevariationfacilitatestheinterpretationofthe incisivesutureinSouthAfricanPlio-Pleistocenehominids.Am.J.Phys. Anthropol.105,121–135.

Braga,J.,Thackeray,J.F.,2016.Kromdraai.AbirthplaceofParanthropusin thecradleofhumankind.SunPress,Johannesburg,SouthAfrica. Braga,J.,Thackeray,J.F.,Dumoncel,J.,Descouens,D.,Bruxelles,L.,Loubes,

J.-M.,Kahn,J.-L.,Stampanoni,M.,Bam,L.,Hoffman,J.,deBeer,F.,Spoor, F.,2013.Anewpartialtemporalboneofajuvenilehomininfromthe siteofKromdraaiB(SouthAfrica).J.Hum.Evol.65,447–456. Braga,J.,Loubes,J.-M.,Descouens,D.,Thackeray,J.F.,Dumoncel,J.,Kahn,

J.-L.,deBeer,F.,Riberon,A.,Hoffman,K.,Balaresque,P.,Gilissen, E.,2015.DisproportionatecochlearlengthingenusHomoshowsa highphylogeneticsignalduringapes’hearingevolution.PLoSOne10, e0127780.

Braga,J.,Dumoncel,J.,Duployer,B.,Tenailleau,C.,deBeer,F., Thack-eray,J.F.,2016.TheKromdraaihomininsrevisitedwithanupdated portrayalofdifferencesbetweenAustralopithecusafricanusand Paran-thropusrobustus.In:Braga,J.,Thackeray, J.F.(Eds.),Kromdraai.A birthplaceofParanthropusinthecradleofhumankind.SunPress, Johannesburg,pp.49–68.

Braga,J.,Thackeray,J.F.,Bruxelles,L.,Dumoncel,J.,Fourvel,J.-B.,2017. StretchingthetimespanofhomininevolutionatKromdraai(Gauteng, SouthAfrica):Recentdiscoveries.C.R.Palevol16,58–70.

Clark,A.,Glanowski,S.,Nielsen,R.,Thomas,P.D.,Kejariwal,A.,Todd,M.A., Tanenbaum,D.M.,Civello,D.,Lu,F.,Murphy,B.,Ferriera,S.,Wang, G.,Zheng,X.,White,T.J.,Sninsky,J.J.,Adams,M.D.,Cargill,M.,2003. Inferringnonneutralevolutionfromhuman-chimp-mouse ortholo-gousgenetrios.Science302,1960–1963.

Clarke,R.J.,2008.LatestinformationonSterkfontein’sAustralopithecus skeletonandanewlookatAustralopithecus.S.Afr.J.Sci.104,443–449.

(14)

Coleman,N.M.,2009.What doprimate hear?Ameta-analysisofall knownnon-humanprimatebehavioralaudiograms.Int.J.Primatol. 30,55–91.

Coleman,M.N.,Colbert,M.,2010.Correlationsbetweenauditory struc-turesandhearingsensitivityinnonhumanprimates.J.Morph.27, 511–532.

Coleman,M.N.,Kay,R.F.,Colbert,M.W.,2010.Auditorymorphologyand hearingsensitivityinfossilnewworldmonkeys.Anat.Rec. 293, 1711–1721.

Dean,M.C.,1986.HomoandParanthropus:similaritiesinthecranialbase andthedevelopingdentition.In:Wood,B.A.,Martin,L.B.,Andrews, P.(Eds.),Majortopicsinprimateandhumanevolution.Cambridge UniversityPress,Cambridge,pp.249–265.

Dean,M.C.,Wood,B.A.,1981.Metricalanalysisofthebasicraniumof extanthominoidsand Australopithecus.Am JPhysAnthropol54, 53–71.

deBeer,G.R.,1937.Thedevelopmentofthevertebrateskull.Oxford Uni-versityPress,Oxford.

Edwards,A.W.F.,Cavalli-Sforza,L.L.,1964.Reconstructionofevolutionary trees.Pheneticandphylogeneticclassiﬁcation.Syst.Assoc.Publ.6, 67–76.

Felsenstein,J.,1988.Phylogeniesfrommolecularsequences:inferences andreliability.Ann.Rev.Genet.22,521–565.

Felsenstein,J.,2004.Inferringphylogenies.SinauerAssociates, Sunder-land.

Freckleton,R.P.,Harvey,P.H.,Pagel,M.,2002.Phylogeneticanalysisand comparativedata: a testand review ofevidence.Am. Nat.160, 712–726.

GarlandJr.,T.,Dickerman,A.W.,Janis,C.M.,Jones,J.A.,1993. Phyloge-neticanalysisofcovariancebycomputersimulation.Syst.Biol.42, 265–292.

Grine,F.E.,2013.ThealphataxonomyofAustralopithecusafricanus.In: Reed,K.E.,Fleagle,J.G.,Leakey,R.E.(Eds.),Thepaleobiologyof Aus-tralopithecus.Springer,NewYork,pp.73–104.

Harcourt,A.H.,Stewart,K.J.,2001.Vocalrelationshipsofwildmountain gorillas.In:Robbins,M.M.,Sicotte,P.,Stewart,K.J.(Eds.),Mountain Gorillas:threedecadesofresearchatKarisoke.CambridgeUniversity Press,Cambridge,UK,pp.241–262.

Hardus,M.E.,Lameira,A.R.,Singleton,I.,Morrogh-Bernard,H.C.,Knott, C.D.,Ancrenaz,M.,UtamiAtmoko,S.S.,Wich,S.A.,2009.A descrip-tionoftheorangutan’svocalandsoundrepertoire,withafocuson geographicvariation.In:Wich,S.A.,UtamiAtmoko,S.S.,MitraSetia, T.,vanSchaik,C.P.(Eds.),Orangutans.Geographicvariationin behav-ioralecologyandconservation.OxfordUniversityPress,Oxford,UK, pp.49–64.

Hughes,A.R.,Tobias,P.V.,1977.AfossilskullprobablyofthegenusHomo fromSterkfontein.Transvaal.Nat.265,310–312.

Kuman, K., Clarke, R.J., 2000. Stratigraphy, artefact industries and hominidassociationsforSterkfontein,Member5.J.Hum.Evol.38, 827–847.

Lieberman,D.E.,Pilbeam,D.R.,Wood,B.A.,1996.Homoplasyandearly Homo:ananalysisoftheevolutionaryrelationshipsofH.habilissensu strictoandH.rudolfensis.J.Hum.Evol.30,97–120.

Marshall,A.J.,Wrangham,R.,ClarkArcadi,A.,1999.Doeslearningaffect thestructure ofvocalizationsin chimpanzees?Anim. Behav.58, 825–830.

Martins,E.P.,1999.Estimationofancestralstatesofcontinuous charac-ters:acomputersimulationstudy.SystematicBiology48,642–650. Moggi-Cecchi,J.,Tobias,P.V.,Beynon,A.D.,1998.Themixeddentitionand

associatedskullfragmentsofajuvenilefossilhominidfrom Sterk-fontein,SouthAfrica.Am.J.Phys.Anthropol.106,425–466. Mohammadi,A.,Jufas,N.,Sale,P.,Lee,K.,Patel,N.,O’Leary,S.,2016.

Micro-CT analysisoftheanatomicalcharacteristicsofthe stape-dialannularligament.Anat.Sci.Int.,1–15,http://dx.doi.org/10.1007/ s12565-016-0331-4.

Noussios,G.,Chouridis,P.,Kostretzis,L.,Natsis,K.,2016.Morphological andmorphometricalstudyofthehumanossicularchain:areviewof theliteratureandameta-analysisofexperienceover50years.J.Clin. Med.Res.8,76–83.

Nunn,C.L.,2011.Thecomparativeapproachinevolutionaryanthropology andbiology.TheUniversityofChicagoPress,Chicago.

Organ,C.,Nunn,C.L.,Machanda,Z.,Wrangham,R.W.,2011.Phylogenetic rateshiftsinfeedingtimeduringtheevolutionofHomo.Proc.Natl. Acad.Sci.USA108,14555–14559.

Pagel,M.,1999.Inferringthehistoricalpatternsofbiologicalevolution. Nature401,877–884.

Pagel,M.,2002.Modellingtheevolutionofcontinuouslyvarying char-acters on phylogenetic trees. In:MacLeod, N., Forey,P.L. (Eds.), Morphology,shapeandphylogeny.CRCPress,BocaRaton,FL,USA, pp.269–286.

Quam,R.,Martinez,I.,Rosa,M.,Bonmati,A.,Lorenzo,C.,deRuiter,D.J., Moggi-Cecchi,J.,CondeValverde,M.,Jarabo,P.,Menter,C.G., Thack-eray,J.F.,Arsuaga,J.-L.,2015.Earlyhomininauditorycapacities.Sci. Adv.1,e1500355.

Revell,L.J.,2012.Phytools:anRpackageforphylogeneticcomparative biology(andotherthings).MethodsEcol.Evol.3,217–223. Sim,J.H.,Röösli,C.,Chatzimichalis,M.,Eiber,A.,Huber,A.M.,2013.

Char-acterizationofstapesanatomy:investigationofhumanandguinea pig.JARO14,159–173.

Strait,D.S., (Ph.D.Dissertation)1998.Evolutionaryintegrationinthe hominidcranialbase.StonyBrook,StateUniversityofNewYork. Strait,D.S.,Grine,F.E.,2004.Inferringhominoidandearlyhominid

phy-logenyusingcraniodentalcharacters:theroleoffossiltaxa.J.Hum. Evol.47,399–452.

Strait,D.S.,Grine,F.E.,Moniz,M.A.,1997.Areappraisalofhominid phy-logeny.J.Hum.Evol.32,17–82.

Wood,B.A.,Constantino,P.,2007.Paranthropusboisei:ﬁftyyearsof evi-denceandanalysis.Yearb.Phys.Anthropol.50,106–132.

Wood,B.A.,Boyle,E.K.,2016.Hominintaxicdiversity:factorfantasy? Yearb.Phys.Anthropol.159,S37–S78.